|
--- |
|
license: other |
|
base_model: "black-forest-labs/FLUX.1-dev" |
|
tags: |
|
- flux |
|
- flux-diffusers |
|
- text-to-image |
|
- diffusers |
|
- simpletuner |
|
- lora |
|
- template:sd-lora |
|
--- |
|
|
|
# simpletuner-lora: Pulp-Fantasy Comics in Frank Frazetta style |
|
|
|
A LoRA for the Flux-1.dev text-to-image model to enforce a vintage fantasy pulp-comic style with high contrast and simple linearts - a style in reminiscent of the old Frank Frazetta fantasy comics. |
|
|
|
The model was trained on 76 manually captioned images in different resolutions - many thanks to **SubtleOne** who gave me the training data. |
|
|
|
## Validation settings |
|
- CFG: `3.5` |
|
- CFG Rescale: `0.0` |
|
- Steps: `30` |
|
- Sampler: `Euler` |
|
- Resolution: `1024x1024` |
|
|
|
|
|
<Gallery /> |
|
|
|
|
|
## Inference |
|
|
|
|
|
```python |
|
from diffusers.pipelines.flux.pipeline_flux import FluxPipeline |
|
import safetensors.torch |
|
pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16, token="hf_[yourhuggingfacetoken]") |
|
weights = safetensors.torch.load_file("frazetta_pulp_comics.safetensors") # download it manually |
|
pipe.load_lora_into_transformer(weights, pipe.transformer) |
|
image=pipe("Illustration by Frank Frazetta: A grand medieval banquet hall illuminated by flickering torchlight, with long wooden tables laden with roasted meats, bread, and wine. Nobles in fine garments and elaborate headpieces feast and converse, while minstrels play lively music in the background. At the head of the hall, a king sits on a grand throne, overseeing the festivities. The hall's stone walls are adorned with tapestries depicting epic battles and legendary heroes, creating an atmosphere of opulence and history.") |
|
image.images[0].save("output.png") |
|
``` |
|
|
